Vertica requesting huge memory

Navin_C

We have recently found an issue where Vertica is requesting more amount of memory for queries.

 

To start with, 

this question on this forum explain Verticá's memory management behaviour.

 

The way vertica handles memory for each query is still an issue for us, but another issue for us why vertica demands more memory for some of our queries.

 

We have few queries which demand more than 4 GB of memory (I think 4 GB is also too much , but its our budget that gives this memory to our queries.)

 

The memory asked by Vertica can be anything above normal (9 GB, 45 GB, 78 GB).

 

We have checked the stats and everything seems to be fine.

 

Also, is there any document, which explain reading explain verbose more clearly.

Navin_C

Adding to my question above,

 

Is this how we should calculate the amount of memory required by any query ?

 

Vertica memory calculations are explained below:

1 MB * Number of cores/node * Number of columns of all tables involved in the query) + ( 2 MB * Number of DVROSs).

For example: 
Say each node in the cluster has 24 cores.
There are 2 tables involved in the query, each table is having 50 columns and there are 20 DVROSs 
(i.e. there are 20 records in the delete_vectors table for the projections related to the query)

(1 MB * 24 * 2*50 + 2MB*10 = 2400 MB) + (2 MB * 20 = 40MB) = 2440 MB = 2.38GB.

We need 2.38 GB/node to process that query.a

 

Is the above calculation true for latest versions.